Output Gap and Model Uncertainty

Output gap estimates play a key role in macroeconomic policy, but different estimation methods often give conflicting readings of the cyclical position. This paper explores the quasi-real-time dynamics of model uncertainty, defined as the cross-method dispersion of output gap estimates. To estimate output gaps and quantify model uncertainty, we use thirteen univariate, multivariate, and institutional output gap measures for 25 EU member states and the United Kingdom over 2002:Q1 to 2024:Q4. The results show an asymmetric pattern: model uncertainty rises sharply in contractions but shows no clear relationship with growth in expansions, where it is also more persistent. A nonlinear dynamic panel model estimates the threshold separating the two regimes at a quarterly GDP growth rate near minus one percent, which matches the recession threshold implied by Friedman’s plucking model of business cycles. Out-ofsample panel probit forecasts that use model uncertainty as the only predictor track recessions reasonably well at horizons up to one year, with the best performance at short horizons. A systematic increase in model uncertainty in recessions complicates the decision-making in real time, especially around the business cycle turning points.